ArxModel train

---

train(self, frame, timeseries_column, x_columns, y_max_lag, x_max_lag, no_intercept=False)

[ALPHA] Creates AutoregressionX (ARX) Model from train frame.

Parameters:

frame : Frame

A frame to train the model on.

timeseries_column : unicode

Name of the column that contains the time series values.

x_columns : list

Names of the column(s) that contain the values of previous exogenous regressors.

y_max_lag : int32

The maximum lag order for the dependent (time series) variable

x_max_lag : int32

The maximum lag order for exogenous variables

no_intercept : bool (default=False)

a boolean flag indicating if the intercept should be dropped. Default is false

Returns:

: dict

A dictionary with trained ARX model with the following keys:

c : float64

intercept term, or zero for no intercept

coefficients : list

coefficients for each column of exogenous input.

Fit an autoregressive model with additional exogenous variables.

Notes

1. Dataset being trained must be small enough to be worked with on a single node.
2. If the specified set of exogenous variables is not invertible, an exception is thrown stating that the “matrix is singular”. This happens when there are certain patterns in the dataset or columns of all zeros. In order to work around the singular matrix issue, try selecting a different set of columns for exogenous variables, or use a different time window for training.

Examples

See here for examples.